|
Activity Number:
|
381
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Quality and Productivity
|
| Abstract - #304994 |
|
Title:
|
Heteroscedasticity and Non-Normality in Regression: A Parametric Approach
|
|
Author(s):
|
Fassil Nebebe*+
|
|
Companies:
|
Concordia University
|
|
Address:
|
Dept. of Decision Sciences & MIS , Montreal, QC, H3G 1M8, Canada
|
|
Keywords:
|
bootstarp ; robust parameter designs
|
|
Abstract:
|
The normal theory of estimation and hypothesis testing in regression may be adversely affected by the heterogeneity and non-normality of the error terms, and is particularly serious for constructing prediction intervals if one or both of these assumptions are violated. Very often, the Box-Cox power transformation of the dependent variable is used to manage the problems. However, difficulties may arise if there does not exist a single transformation that rectify both problems. We suggest a parametric approach for a more direct analysis using a class of skewed error distributions arising from the approximation of marginal distributions in importance sampling. The likelihood test of normality is discussed and compared with some popular normality tests. The efficiency of the present approach in estimation and prediction interval construction is demonstrated in Monte Carlo studies.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2009 program |
